Irrigation efficiency paradox

Why can upgrading a valley's farms to efficient drip irrigation leave less water in the river, not more?

A valley floods its furrows, and most of what leaves the canal never touches a root. Replace that with drip lines and every farm can show you the meter: less water applied per hectare, verifiably. Yet a decade on, the river below carries less, the wetland at its mouth is contracting, and the water table is falling. Efficiency subsidy is one of the most popular water policies in the world and rests on an inference so obvious nobody checks it, so it is worth finding precisely where that inference breaks.

Start with a phrase doing too much work. "Water used" names two quantities, and the whole puzzle lives in the gap between them. One is withdrawal: what you divert. The other is depletion: what leaves the basin's reachable stock for good. Which does an efficiency figure measure?

Follow a hundred units through a flooded field. Thirty are transpired by the crop — breathed out as vapour, genuinely gone. Ten evaporate from wet soil and open canals, also gone and doing nothing useful on the way. The other sixty percolate past the root zone into the aquifer or run off the field's tail back into the river. Where are those sixty?

Not lost. That is the move the intuition misses. Return flow is somebody's supply: the farm below, the town's wells, the river's dry-season baseflow, the marsh. Water law knows this even where engineering forgets it — downstream entitlements in mature basins are quantified against historic return flows, because those flows are what the downstream user has always actually received.

So the flood field's "efficiency" of thirty per cent describes how much of the diversion the crop caught on the first pass. It does not describe how much the basin lost. The basin lost forty.

Now install drip on the same hectare. Divert forty. Transpiration rises — call it thirty-six — because a plant kept steadily moist is never water-stressed and never stops growing, and more transpiration is what a higher yield largely consists of. Evaporation falls to about one, since the soil between emitters stays dry. Return flow: three. Field efficiency is now ninety per cent.

Work the ledger. Depletion has fallen from forty to thirty-seven — a real saving of three, drawn entirely from the evaporation that was doing nobody any good. Return flow has fallen from sixty to three. The downstream user has lost fifty-seven so the basin could gain three. (The numbers illustrate the accounting; they are not a measurement of any real valley.)

That is the first mechanism, and it is pure accounting: efficiency converts recoverable water into unrecoverable water. It does not create water.

The second is larger. Our farmer still holds a right to divert a hundred, is not obliged to return the sixty he no longer needs, and under "use it or lose it" doctrines is actively punished for trying. At forty a hectare he can now irrigate two and a half hectares on the same right. Depletion: ninety-two and a half. Return flow: seven and a half. Every farm is more efficient than before and the valley now consumes more than twice what it did.

And the expansion ratchets, because drip economics favour perennials — orchards, vines — and a perennial cannot be fallowed in a drought without destroying the capital. The basin loses its flexibility along with its water.

The claim is not that efficiency never saves water. It saves precisely the non-recoverable losses — evaporation from wet soil and open canals, and water that would have drained to the sea or into a saline aquifer nobody can pump. That share is real and sometimes substantial. The error is applying a field-boundary ratio to a basin-scale question, so that recoverable losses get counted as savings.

Two folk explanations do not survive this. "The farmers wasted the savings out of greed" fails because the return-flow collapse happens even if not one hectare is added — the hydrology alone does it. And "drip must not work as well as claimed" fails in the opposite direction: the field-scale gains are real and measured. The technology is not the problem; the accounting boundary is.

The test the account must pass: it predicts the paradox appears only where return flows were recoverable. In a basin whose irrigation drains straight to the sea or into brine, efficiency should genuinely increase upstream availability — and does. It predicts too that efficiency paired with a binding cap on consumption, or a retired share of the right, leaves water in the river. A case with recoverable returns, no expansion of area, and a river that gained anyway would break it outright.

Pushing back on myself: the arithmetic is solid, the empirical share is not. Basins that modernised also intensified for other reasons — new markets, cheap pumps, a run of dry years — and separating the efficiency channel from those confounds is genuinely contested. Grafton and colleagues put the argument in Science in 2018 and it has not been overturned, though the magnitude is still argued (recalled). I will not give a global figure for depletion caused by efficiency programmes, because none exists cleanly.

• Why capping consumption rather than diversion is administratively hard, and what measuring it would take.
• How groundwater pumping decouples a farm from its surface right, and what that does to this ledger.

Every term the collection defines is gathered in the glossary.